발명의 언어를, 세계의 언어로.

Referring to FIG. 4, the method for selecting and predicting genes related to biological characteristics according to one embodiment of the present disclosure includes: obtaining single-cell gene expression profiles containing gene expression patterns of multiple single cells from cancer tissues S101; generating gene correlation information by clustering a plurality of genes associated with co-expression characteristics through network analysis of the obtained single-cell gene expression profile S103; determining gene importance for each gene in the first cancer as a target using a first machine learning model S105; generating second training data for the task of the target based on the determined important genes for the target S107; performing the task of the target using the second training data, which includes information about the determined genes S109; and re-determining at least some of the genes determined from the performed task, and re-performing the task to predict the target based on information about the re-determined genes S111.

The method of claim 5, wherein generating the gene correlation information comprises: producing correlation coefficients for gene pairs based on the single-cell gene expression profiles; creating a weighted network based on the produced correlation coefficients; and classifying co-expressed genes into gene modules by grouping nodes in the created network using clustering algorithms.